Editor's note: This is the second of a two-part series on the challenges of measuring return on investment for generative artificial intelligence. In the first part, CFO Dive described how finance executives are pouring billions of dollars into generative AI without reliable earnings estimates.

For institutional investors, "speculation" is an unseemly word—unless they happen to profit from it. For finance executives, acting on speculation is often a recipe for losing their jobs.

Yet CFOs across multiple industries are pouring billions of dollars into generative AI based on vague earnings forecasts.

"It's a noisy estimate," said Daniel Rock, co-founder of AI consulting firm Workhelix, referring to common revenue forecasts for generative AI. "But that doesn't mean the attempt isn't worthwhile."

Generative AI has already helped CFOs fine-tune customer service, optimize forecasting, accelerate software updates, and upgrade other tasks.

However, as AI brings about the "industrialization of knowledge production," finance executives face challenges in accurately predicting the ROI of powerful new initiatives aimed at extracting more value from data, noted Laura Veldkamp, a finance professor at Columbia Business School.

This challenge often breaks the conventional standards for measuring ROI: data is inherently harder to observe and price than the assets that underpinned the industrial economy era, such as buildings and workers, Veldkamp said at a productivity symposium held by the Federal Reserve Bank of New York in February.

"There are many things in the economy that are hard to measure, and we don't know enough about them yet," said Prasanna Tambe, an associate professor at the University of Pennsylvania's Wharton School, at the same symposium. "Measurement remains a challenge."

Despite the uncertainty, CFOs and AI experts say the surge into generative AI has swept up investors, finance executives, and information technology companies of all sizes.

"Speculative fervor is part of technology, so there's no need to fear it," said David Cahn, a partner at Sequoia Capital, in a June research report. He observed a $600 billion gap between corporate spending on AI-related infrastructure and the revenue needed to justify that spending.

"Those who stay calm at this moment have the opportunity to build extremely important companies," Cahn said. "But we need to make sure we don't believe the illusion that has spread from Silicon Valley to the nation and the world—the illusion that because AGI (artificial general intelligence) is coming tomorrow, we'll all get rich quickly."

Finance executives willing to take risks on generative AI can follow these seven strategies to cut through the forecasting fog and capture potential returns:

1. Be flexible in evaluating ROI

The hype around any "new new thing" requires CFOs to remain skeptical. However, when a potentially game-changing technology emerges, top finance executives should consider being more flexible in evaluating ROI.

"As a CFO, the first thing you say to someone coming to ask for money is: 'What's the ROI?'" said Glenn Hopper, CFO of Eventus Advisory Group, in an interview. "For something like AI, it's hard to pin down."

He noted that efficiency gains are often one of the most direct returns. AI can shorten the time between delivery of goods or services and payment (i.e., days sales outstanding) by up to three days, streamline just-in-time inventory control, and avoid errors.

"Without prior knowledge, it's difficult to assess potential investment returns," said Gregory Daco, chief economist at EY-Parthenon, in an interview. Precisely identifying potential returns for each company can eliminate some of the ambiguity.

"We work with many clients case by case, saying: 'In your specific industry, for a company with X market position and X total addressable market, we think if AI is applied to these functions, it could generate this kind of return,'" Daco said. "It really is case by case because every industry and company is different."

Rock said in an interview that many companies succeeding with generative AI embrace risk and experiment, despite uncertain ROI.

"If it fails, it doesn't necessarily mean you made a bad choice," Rock said. "You can terminate it and pivot in another direction." Rock co-founded Workhelix with Erik Brynjolfsson and Andy McAfee.

A June survey by KPMG found that C-suite executives have shown flexibility this year in their ROI expectations for generative AI.

In the first quarter, 51% of business leaders expected the biggest gains over the next 12 months to come from productivity improvements, while 47% expected revenue growth to be the primary benefit, KPMG said.

In the second quarter, the rankings reversed, with 52% of the 100 respondents listing revenue growth as the biggest improvement, while productivity gains fell to 40%, according to KPMG data.

2. Focus on long-term gains

Rock said launching generative AI typically requires significant upfront investment that could otherwise be used for businesses with more certain and immediate returns.

"There's an opportunity cost to building intangible things, like new ways of doing things, new organizational structures, or new workflows. It looks like you're putting in a lot and getting nothing," he said. "Later, you think, 'Hey, we're much more efficient'—and you start seeing tangible gains."

Paradoxically, CFOs who know AI best—its capabilities, limitations, and costs—often expect returns to materialize within three to five years, Daco said. Peers with less knowledge of the technology "tend to believe returns will be more immediate—within two years."

3. Develop a "holistic" strategic plan

Daco said that before considering potential investment returns, CFOs should help develop a strategic plan to streamline operations and drive growth by integrating generative AI into the organization. Simply bolting the technology onto existing operations often yields little benefit.

"It's not just about saying 'we can automate or enhance certain functions,'" he said. "It's about taking a holistic view of how to integrate AI into the organizational architecture, rather than a simple add-on."

As with any strategic plan, setting goals from the start is crucial, said Adriana Carpenter, CFO of expense management software provider Emburse. "It's best to start with the question: 'What outcome do I want to achieve?'" she said.

While developing broad strategy from a top-down perspective, CFOs should also ensure that employees, down to the front line, can fully leverage generative AI, CFOs and AI experts said.

"You need to cultivate an AI-literate corporate culture and think about how to optimize the workforce so you can combine a top-down strategic perspective with a bottom-up operational perspective," Daco said.

4. Prioritize AI projects

CFOs can apply generative AI to most corporate functions but should resist the temptation and prioritize, finance executives and AI experts said.

"You may need to rank various opportunities because there are many, and resources are certainly limited," Rock said.

Beyond alignment with strategy, generative AI projects should leverage existing employees, focus on operations they find attractive, and offer the prospect of quick wins, he said.

Before launching a proof of concept, "pick a project you know will succeed and change minds," Rock said. "Once you've shown ROI, you can say: 'Okay, now we can move forward.'"

Some early adopters of generative AI, while ultimately confident in achieving company-wide benefits, first focus on picking a safe project rather than risking failure that could turn employees against the technology, Rock said.

"You have to prioritize the business outcomes you want to achieve," Carpenter said.

Generative AI is particularly effective at boosting revenue by analyzing customer data and monetizing it, Carpenter said. It can also help identify where a company is underpricing relative to competitors or how to expand its customer base.

5. Build friendly relationships with CTOs and CIOs

As the keeper of the purse strings, CFOs may have tense relationships with chief information officers or chief technology officers, who may not always secure funding for their top projects, AI experts and finance experts said. CFOs should set a friendly, collaborative tone.

"CFOs and CIOs—often—their domains have historically been seen as cost centers, so they compete with each other," Hopper said. "It's important to collaborate to improve efficiency and truly understand that AI is a strategic capability that enhances a company's financial performance and operational excellence—if they work together, everyone wins."

CFOs also need to understand technology more deeply than in past decades, AI experts and CFOs said.

Previously, CFOs primarily focused on how AI and automation could help with financial controls, back-office, and other finance functions. Now, CFOs need to broaden their perspective and understand the value of generative AI across the entire company.

"I'm not saying you have to become a data scientist or developer, but you need to understand your business, your industry, to have that breadth of knowledge," Hopper said. "If you don't know how information is created, how can you correct it?"

Tech-savvy CFOs will also more quickly identify high-return innovations in generative AI.

"Every week there are new breakthroughs—there's always something novel," Rock said. "You need to keep up with new developments and build your own detectors to spot innovations that are highly valuable to the company."

6. 'Define' data

As part of accurately measuring ROI, CFOs need to ensure that data fed into generative AI is precisely defined across the company, AI experts and CFOs said. Otherwise, different parts of the company may use the same data to reach inconsistent conclusions, undermining the technology's value.

"Early on, you'll encounter data silos because everyone is doing their own citizen development, building their own data fiefdoms," Hopper said. "If you only take a piece of the pie each time, you won't get maximum value. The way to get maximum value is through a shared data warehouse used across the company."

CFOs should remember that the costs of launching generative AI applications—including creating a "data dictionary" and detailed information governance as part of "responsible AI"—may exceed estimates, Carpenter said.

7. Engage employees

CFOs seeking high ROI will make a mistake if they primarily use generative AI to cut payroll, AI experts and CFOs said. Instead, the technology offers a potential tool to streamline operations and boost employee job satisfaction.

"Viewing this technology as a way to save labor costs rather than as an enhancement tool is a huge mistake. It enables people to accomplish a great deal of work and focus on truly difficult things," Rock said.

Large language models like ChatGPT will put jobs in dozens of occupations at risk, including accountants, auditors, financial quantitative analysts, blockchain engineers, interpreters, mathematicians, and journalists, according to a study by researchers at the University of Pennsylvania and OpenAI.

The researchers said the technology will gradually simplify at least 10% of tasks for 80% of workers, and half of tasks for 19% of workers.

To ease employee anxiety, CFOs should "cultivate a culture of AI literacy that is eager to move forward," Daco said. They should ensure employees "support the transformation and alleviate fears of job loss."

Finance executives and AI experts said precisely measuring the ROI of generative AI will not happen within months. To succeed, CFOs need to balance determination with patience.

"There's a lot to learn, there will be iterations, and every business in different industries will take time to evolve and accelerate adoption," Carpenter said. "But generative AI is here to stay."